AI at the Edge, Part 2: Operationalizing Intelligence at Scale
The Skinny
-
Edge AI is enabling tailored, policy-driven video intelligence, allowing organizations to define and enforce scenario-specific security rules in real time.
-
Solutions like Camio and Deep Sentinel demonstrate how edge computing scales, supporting both complex enterprise deployments and smaller, budget-conscious environments.
-
Built-in privacy, compliance and data governance capabilities are becoming essential features of modern edge AI security platforms.
In Part 1 of this series, we examined how edge AI is enhancing real-time responsiveness in video surveillance and access control. Now, we continue the journey with more cutting-edge solutions built for flexibility, affordability and enterprise-grade performance.
(Editor’s note: This article is an installment of the “Real Words or Buzzwords?” series about how real words can become empty words and stifle technology progress.)
Camio: Video intelligence AI for scenario-specific needs
Camio offers a uniquely flexible approach to video intelligence by enabling organizations to create use plain text to describe policies they want monitored and create do-it-yourself (DIY) AI models tailored to their own operational scenarios and policies. Rather than relying on pre-packaged analytics tuned for general purposes, Camio empowers customers to define exactly what activity is important to detect, monitor and respond to, allowing security teams to customize detection criteria for the specific environments they protect.
Camio’s platform leverages large language models (LLMs) and large multimodal models (LMMs) to generate ongoing situational narratives that describe activity within an area under surveillance. This enables real-time awareness not just of isolated events, but of evolving context over time, providing a richer operational picture than traditional alert-driven systems.
Multiple cameras and audio sensors can be logically linked to define an indoor or outdoor area of interest. For that area, security teams can input security policies, standard operating procedures (SOPs) and guard post orders as contextual data. This enables Camio’s AI to interpret sensor input accordingly and support appropriate response actions automatically. When implementing this functionality, it’s important to collaborate with the organization’s data governance and compliance stakeholders. Some organizations are required to document their use of video-based PII, including images of bodies and faces, location data tied to individuals, and vehicle identifiers such as license plate numbers, signage or branding and visible occupants.
Camio supports “starting small” and expanding its application, as is sometimes required to work within budgetary constraints. Many Camio customers have found additional use cases and either added more cameras or begun applying its AI to more cameras. Camio’s edge AI software runs onsite in virtual machines (VMs) and can operate on single or multiple physical servers. If an unplanned expansion requires additional edge computing, another server can be added if the existing server’s capacity is insufficient.
Uniquely, AI edge computing is a factor in the Camio subscription pricing. On-site initial processing takes a processing load off the Camio cloud platform, and this cloud cost savings is passed on to the end user.
While AI edge computing is preferable from two perspectives — performance and total cost of ownership (TCO) — Camio does not require edge hardware. Video feeds can be ingested directly from video management systems (VMS) or directly from IP cameras, allowing Camio to support the full range of deployment architectures and existing security system capabilities. Camio also accepts video and image clips via direct upload and email, in programmatic APIs, providing a versatile approach for video evidence intake and system integration.
Similar to Actuate, when edge AI systems are used to send alerts to Camio, it’s important that they be tuned for zero false negatives — especially considering any DIY scenario-specific scope that Camio is applying. Overall, Camio’s combination of DIY AI modeling, LLM-driven contextual narratives, flexible integration options and distributed edge/cloud architecture delivers a solution designed for scenario-specific video intelligence. This enables organizations to operationalize security policies dynamically and at scale.
Deep Sentinel: Enterprise-grade surveillance for smaller sites
Deep Sentinel delivers AI-powered remote guarding capabilities combined with site surveillance and response technology that were once only practical for Fortune 500 companies. It overcomes the cost and complexity barriers of earlier high-end systems while avoiding the design and capability constraints that have often limited small business and residential solutions.
Recent advances in security technology, coupled with game-changing AI capabilities, have made that possible, enabling a remote guarding service tailored for residential homes, apartments, small-to-medium businesses, and commercial and retail environments.
At its core is an on-premises edge computing and connectivity Hub: built using an AMD or Intel NUC mini-computer, a small form factor computer that represents a significant advancement in compact, high-performance computing. The Hub connects cameras and other devices and is a point of device integration. It runs all the AI at the edge, analyzing live video feeds from cameras in real time.
Once a potential threat is identified, the system streams the event to a live specially-trained security guard who can engage directly using two-way audio, activating built-in sirens or even deploying FlashBang deterrent options (e.g., smoke bombs, pepper spray, strobe lights) to actively stop intruders.
Deep Sentinel also provides physical duress buttons for what it calls its SentinelNow service. When an employee feels unsafe, observes suspicious activity or simply needs an extra set of eyes before they walk to the dumpster or parking lot, they can press the SentinelNow button to instantly connect with a Deep Sentinel guard.
Within seconds, the guard views the live camera feed, provides verbal intervention or support when appropriate and escalates to emergency services immediately when needed. In this emergency scenario, the guard can then provide first responders with real-time situational awareness and details such as number and description of suspects, any visible weapons, crime details and any known medical needs.
Deep Sentinel supports third-party camera partner program, called a Bring Your Own Camera (BYOC), enabling customers to connect existing IP or PoE cameras, ensuring seamless integration with current systems.
The Hub’s local AI inferencing enables ultra-fast threat detection and alert verification, minimizing false alarms and reducing bandwidth use. It delivers real-time, live guard intervention and deterrent action as a unified, enterprise-capable surveillance solution that’s accessible to smaller sites.
The Deep Sentinel Gen V Hub typically supports up to 20 cameras, providing flexibility for both residential and small enterprise deployments. This design prioritizes low latency, proactive deterrence and consistent operation even in smaller-scale settings.
Deep Sentinel’s stated objective is to transform how people experience safety, by providing a platform that delivers real-time intervention at the highest standard in the industry. Underscoring that commitment is the response time profile defined in its business and residential Service Level Agreements (SLAs):
Deep Sentinel defines its response time in three phases: AI detection occurs within approximately 5-7 seconds; video streaming to a live guard takes about 2-8 seconds; and guard intervention typically begins within 30 seconds of the event for 10 common scenarios in both business and residential settings. For certain after-hours events — such as trespassing, auto-loitering and door loitering — intervention is expected within 60 seconds.
Guards follow a four-tier escalation system:
- Dismiss harmless activity
- Continue to Observe unknown behaviors
- “Hello” (Trust but Verify) for potentially suspicious actions
- “Stop” (Criminal Intent) for clear threats
The design of Deep Sentinel’s service demonstrates the role AI edge computing can play in enabling high-speed, scenario-appropriate, technologically automated and human-guided response.
Compliance, privacy and data governance concerns
AI edge computing security devices, given their processing power and the criticality of the data they process, are increasingly targeted by advanced cyber threats. For large enterprises, this underscores the need for data privacy, SOC2/ISO compliance, auditability and security system hardening as foundational requirements — not nice-to-haves. Recent IoT attacks, like the 2022 Mirai botnet resurgence, highlight these risks. Aligning with organizational data governance and compliance policies helps ensure that physical security AI edge computing’s performance benefits are not compromised.
Accounting for expansion
Platforms like Ambient.ai, Camio and Deep Sentinel have differing edge computing requirements and options. It’s important to understand how the edge computing component can scale when the camera deployment expands. For Deep Sentinel, it’s simple — just add another Hub when the camera count grows beyond an increment of 20. Where edge servers are involved, such as with Ambient.ai and Camio, there are more design and cost-factor tradeoffs to consider.
AI edge computing and high-caliber security response
Getting the edge-cloud alignment right drives smarter, faster security responses. The solutions reviewed above show that AI edge computing is indeed transforming physical security response capabilities. So, it’s safe to say that when it comes to AI for physical security, edge computing is much more than a buzzword.

Ray Bernard, PSP, CHS-III
Ray Bernard, PSP CHS-III, is the principal consultant for Ray Bernard Consulting Services (www.go-rbcs.com), a firm that provides security consulting services for public and private facilities. He has been a frequent contributor to Security Business, SecurityInfoWatch and STE magazine for decades. He is the author of the Elsevier book Security Technology Convergence Insights, available on Amazon. Mr. Bernard is an active member of the ASIS member councils for Physical Security and IT Security, and is a member of the Subject Matter Expert Faculty of the Security Executive Council (www.SecurityExecutiveCouncil.com).
Follow him on LinkedIn: www.linkedin.com/in/raybernard.
Follow him on Twitter: @RayBernardRBCS.